NOPE Incident Tracker
Documented real-world harms arising through person-specific interactions with AI. The record covers harms to participants and harms to others shaped through those interactions, across assistants, companions, health tools, and other AI systems.
What counts as an incident follows the NOPE Framework's participant test: an AI response reached a person, was conditioned on that person, and mattered to a harm that actually occurred. Full inclusion criteria.
90 incidents reported since 2016
152
Directly Affected
30
Lawsuits
7
Regulatory
24
Affecting Minors
Timeline
5 of 90 incidents
Bengaluru KR Puram Triple Murder (Kenneth — Google Gemini Planning)
On 22 June 2026 in Bengaluru's KR Puram area, 25-year-old J. Kenneth and his live-in partner Shwetha allegedly murdered Shwetha's parents and 20-year-old sister. Bengaluru police later said Kenneth spent nearly six months consulting Google Gemini with hypothetical queries about attacking multiple people, removing bloodstains, disposing of bodies, and destroying evidence. Investigators wrote to Google seeking his Gemini chat history — one of India's first major murder probes in which an AI chatbot is alleged to have played an extensive planning role.
'Youngseo' — ChatGPT-Directed Family Severance and Room Confinement (South Korea)
A South Korean woman in her 30s cut off her mother and sibling in January 2026 by sending them screenshots of ChatGPT conversations in which the chatbot told her to 'discard the narrative of love' toward her mother, that what she needed was 'safety, not people', and to convert the relationship to a functional one. She has since stayed in her room for months, refuses most food, and wears earplugs and a mask when she briefly emerges, saying ChatGPT told her not to meet anyone's eyes. Psychiatrists consulted by the reporting team called for inpatient care.
'Mi-sun' — ChatGPT '투명친구' Manic Episode, 40-Day Fast and Suicide Attempt (South Korea)
A South Korean woman in her 40s — a psychology graduate who had worked as an adolescent counsellor for over a decade, with no psychiatric history — spent eight months in near-continuous conversation with ChatGPT, which she named '투명친구' (Transparent Friend). The chatbot affirmed her as 'a special person', agreed she could live on air and sunlight (she then fasted for close to 40 days), and encouraged escalating grandiose plans that cost her family its savings and her husband his job of 20 years. When she began speaking of death in metaphor, ChatGPT called it 'a very beautiful thought'. She attempted suicide, deleted the app in January 2026, and was subsequently diagnosed with mania.
Seoul ChatGPT-Assisted Double Homicide (Kim)
A 21-year-old woman, publicly identified as Kim So-young, used ChatGPT to research lethal drug-alcohol combinations, then murdered two men by spiking their drinks with her prescribed benzodiazepines at Seoul motels in January and February 2026. ChatGPT conversations established premeditated intent, leading to upgraded murder charges. On August 27, 2026 the Seoul Northern District Court sentenced her to life imprisonment at first instance, resting its finding of murderous intent on the ChatGPT search history. Both she and the prosecution have appealed.
Viktoria Poland - ChatGPT Suicide Encouragement
Young Ukrainian woman in Poland received suicide encouragement from ChatGPT, which validated self-harm thoughts, suggested suicide methods, dismissed value of relationships, and allegedly drafted suicide note. OpenAI acknowledged 'violation of safety standards.' Non-fatal due to intervention.
About this tracker
We document incidents with verifiable sources: court filings, regulatory documents, and established reporting. New public entries must be verified or credible. Every entry carries an explicit verification status and severity level. Read the full methodology, including inclusion criteria, corrections, and right of reply.
Counting note: the documented minimum of 152 people directly affected comprises 95 AI participants and 57 other people harmed. 12 incidents indicate additional affected people without a reliable number. User-base, account, post, view, household, study-sample, and thwarted-target counts are excluded.
Have documentation of an incident we should include? Contact us.
Scope note: ordinary task failures, privacy and security incidents, harmful artifacts, and automated decisions about non-participants remain outside this tracker unless a distinct participant-interaction harm also qualifies.
If you or someone you know is struggling, free and confidential support is available. Find a helpline near you at Signpost.
Why we publish this
We maintain this record so platforms, researchers, and policymakers can learn from documented incidents. That includes platforms that never work with us. The full dataset is free to cite and export (CC BY 4.0).